Placebo controls are a mainstay of clinical research in medicine. They are one of the features imparting gold standard status to the iconic “randomized, double-blind, placebo-controlled clinical trial.” This is surely a recognition of the potency of the placebo effect. Why else would designers of clinical trials be obsessed with placebo controls? In this article, I hope to explain why the value of placebo controls is not an affirmation of a powerful placebo effect.
The primary value of placebo controls in clinical trials is the minimization of bias.
The Rise and Fall of the Placebo Effect
A placebo is an intervention intentionally devised to be devoid of relevant physiologic effects. The apparent over-performance of placebo groups in some studies created the impression of a potent therapeutic effect from these intentionally impotent interventions. This apparent response to an inert intervention has been labelled the “placebo effect” and granted a reputation as a mysterious phenomenon with great scope and power.
More careful studies, and more critical analysis have taken much of the mystery and potency out of the placebo effect. A variety of biases and statistical artifacts can conspire to create the illusion of a response to the placebo. When carefully controlling for these variables, the placebo effect virtually disappears for objective measures. Objective measures are variables that the investigator can measure, such as survival, blood sugar, and temperature. Responses that are reported by the study participants like pain and anxiety are vulnerable to influences that are hard to measure. Subjective responses like these remain the last hiding place for the placebo effect. The neutering and demystification of the powerful placebo narrative has been a frequent subject on SBM, as discussed by Dr Novella and by Dr Gorski in a refutation of a NY Times editorial.
Placebos and clinical trials
Not every research question in medicine is appropriate for a placebo-controlled trial. Such a design may be impossible, impractical, or unethical for certain questions. But when it can be included in the design of a trial, it contributes valuable protection against bias in the conduct of the study. For the remainder of the discussion I will stick to the best-use example of placebo-controlled studies: investigating the safety and efficacy of an experimental medication.
If so-called placebo effects do not influence an outcome we need to measure, why bother with a placebo control in a clinical trial? Let’s break down the components of a randomized, double blind, placebo controlled clinical trial and explain how design features individually and collectively protect against bias.
Bias in clinical trials may be described as systematic errors that encourage one outcome over others. The potential effect of bias is that investigators will come to the wrong conclusions about the beneficial and harmful effects of interventions. Several mechanisms may bias clinical trials, affecting the estimated intervention effects.
Bias is the enemy of sound research. It can never be eliminated completely, but good researchers understand potential sources of bias, neutralize them to the extent possible, and acknowledge other potential sources in their analysis and reporting of data.
Different biases can contaminate the results in different ways. In one situation biases could simulate or exaggerate a therapeutic effect. In another situation biases could minimize or hide a therapeutic effect. Assessment of the harmful effects of a treatment are also highly sensitive to bias.
Mitigation of bias in randomized, double-blind,
placebo-controlled trials
The value of a control group
We want to test the effectiveness of our new drug, humbly called “Panacea” for condition X. We find 1000 patients with condition X and treat them with Panacea. At the end of the study our measurements of condition X are unchanged. Can we conclude that Panacea is ineffective for condition X? Maybe not. If condition X is a relentlessly progressive condition, something like amyotrophic lateral sclerosis (ALS), no change of condition X might be a therapeutic victory. On the other hand, if condition X is self-limited and usually improves spontaneously, like singultus (hiccups!), no change in the condition might be an indication that Panacea is actually making the condition worse.
In order to figure out if a treatment is having a positive or negative effect, we need some reference for comparison. The reference should represent what would be expected of our participants if they did not receive the study drug. There are many possible comparison groups we could imagine, but the ideal control group would be identical to the treatment group except for the study drug.
The value of randomization
There is no perfect control group but, when possible and practical, using a random process to sort experimental and control groups minimizes the risk that the two groups will differ in significant ways.
In a randomized clinical trial, each potential participant is first screened to determine their eligibility. They are informed that they will be randomized into a treatment group or a control group. If a potential participant is eligible and agrees to the terms of the study, they sign a consent form and are enrolled. Only AFTER the participant is enrolled are they randomized to the treatment or control group.
A confounder is a factor, other than the experimental treatment that influences a study participant’s clinical responses relevant to the study. Randomization is likely to balance known confounders such as age, gender, baseline disease severity, etc. But the real power of randomization is that it will balance unknown confounders as well. If some proportion of a population has an undiscovered genetic mutation that predisposes to a more severe form or more rapid progression of disease X, randomization is likely to sort the mutation so it is equally represented in the treatment and control groups, minimizing the chance that this mutation will disproportionally influence the drug or the control group. This is a very powerful advantage! Randomization balances known and unknown confounders, minimizing the potential bias
Randomization enhances the likelihood that treatment and control groups will be similar when they enter the study.
The value of blinding
“Blinding” describes a study design in which knowledge of treatment group (drug vs control) is withheld from relevant individuals involved in the conduct of the study.
Imagine that you are a participant in a clinical trial of an experimental drug. Now imagine that you know you are in the treatment group. You might be more enthusiastic about participating in the study. You might be more likely to comply with study activities. You also might be more likely to report every headache, cough, sleepless night, etc. as a potential drug side effect.
Now imagine that you know that you are in the control group. You might be more concerned that your condition is worsening. You may be more likely to access treatments outside the study protocol. You might be more likely to drop out of the study. You might be less likely to report life events as side effects.
These differences in perceptions and behaviors are likely to have asymmetric influences in important study results. These asymmetric effects are forms of bias.
Designing a study so that participants are not aware of treatment (or lack of treatment) they are receiving is known as “blinding.”
Investigators are not immune to bias. Knowledge of a participant’s status as drug or control might influence the way an investigator interacts with those participants. It might influence the way they interpret findings, and decisions they make during the conduct of the study. It might influence what is collected as an adverse event. It might allow the investigator to accidentally “unblind” the participant to their treatment group. When possible, it is desirable to “blind” the investigator to the assigned treatment group of each participant. A study for which the study participants AND the investigators are unaware of treatment assignment is referred to as “double blind.”
Blinding is a powerful tool to minimize bias during the conduct of a study.
The value of a placebo control
It should be evident that without a placebo control, it is impossible to keep the participants blinded to their assignment. A control participant in a drug trial cannot be blinded to their group assignment if they are aware of whether or not they are taking the study drug. The placebo is designed to resemble the treatment as closely as possible to maintain the blinding of the participants. “Sugar pill” is a term sometimes used to express the inert nature of a placebo. A placebo pill will generally look identical to the study drug. It often contains the same inactive ingredients as the study drug.
Conclusion
Randomization ensures that the treatment and control groups are as similar as possible when they enter the study. Blinding (including the use of placebo controls) ensure that the treatment and control groups are as similar as possible during the conduct of the study. All of these measures are intended to minimize bias so that the study drug can be assessed as fairly as possible. Even if the placebo effect is irrelevant to the drug and the disease being studied, a placebo control remains a powerful tool to maximize the validity of a clinical trial.
